Pipeline: add an optional 'if' and 'after' parameters for conditional tasks and better control flow

[ljgray]

if

Useful for templates pipelines where we might want to skip a task under certain conditions, without messing up key inheritance. The pipeline will internally replace the conditional task with a NoOp if the conditional is False, in which case data keys are simply passed through. Naturally, this will fail for a task which produces a different number of outputs than it takes as input, or which does not intend to pass through the input, so this can likely be made to be more robust.

after

Effectively just streamlines the WaitUntil task - adds a set of data product keys (similar to in, out, and requires) which indicate data products that must exist before a task can run, but are not actually consumed by the task.

IO

Adds two extra tasks:

• Simple task to load all files from a list at once
• More complicated task to load batches of files together. This ends up being necessary because the pipeline never seems producer-only tasks as being "available", so they are simply run in order when there is nothing else to do. However, the order always starts at the first possible task, so two back-to-back IO tasks will result in the pipeline first loading every file for the first task before proceeding.